Mashable recently published an article on a newly launched communication tool that pairs the evolving sector of speech-recognition technology with machine translation (or as we commonly refer to it, “MT”). The result? A new whiz-bang auto-translator that will facilitate communication across language barriers.
While we agree that machine translation is becoming a more viable translation medium (see our Q1 2012 newsletter article for more info on when and how we recommend using MT), language is a constantly evolving and often illogical medium. A word like “suit”, for example, can have several different meanings depending on the context in which it is used. Compounding this issue, MT is notoriously finicky about having clearly set rules of engagement, making logical mistakes a very real challenge. Until machine translation technology matures to the point where it can effectively navigate the fluid nature of language, there will always be a risk of mistranslation.
From our perspective, we feel that the safest and most effective communicator of language, as we know it today, is still a human being. For that reason, when we implement MT solutions at Acclaro, we always recommend human input to make sure that the possibility of error is as low as it can be. The best places to do this are at the beginning of the MT process, where a linguist trains the MT engine on what and what not to output, and at the end, to catch egregious errors before they are released. Our experience has also taught us that MT is most effective on very straightforward content. Creative copy (such as marketing translations) or content where safety and security are concerns are not good MT candidates.
That said, we also agree that MT technology is changing and improving, and nobody really knows what the future will bring. If MT technology can one day rival human output, this tool may be a very effective one.
What’s your opinion? Comment below!
